THE ISLAND DILEMMA: LESSONS OF MODERN BIOGEOGRAPHIC STUDIES FOR THE DESIGN OF NATURAL RESERVES JARED M. DIAMOND Physiology Department, University of California Medical Center, Los Angeles, California 90024, USA ABSTRACT A system of natural reserves, each surrounded by altered habitat, resembles a system of islands from the point of view of species restricted to natural habitats. Recent advances in island biogeography may provide a detailed basis .for understanding what to expect of such a system of reserves. The main conclusions are as follows: The number of species that a reserve can hold at equilibrium is a.function o fits area and its isolation. Larger reserves, and reserves located close to other reserves, can hold more species. If most of the area of a habitat is destroyed, and a fraction of the area is saved as a reserve, the reserve will initially contain more species than it can hold at equilibrium. The excess will gradually go extinct. The smaller the reserve, the higher will be the extinction rates. Estimates of these extinction rates for bird and mammal species have recently become available in a few cases. Different species require different minimum areas to have a reasonable chance of survival. Some geometric design principles are suggested in order to optimise the.function of reserves in saving species. INTRODUCTION For terrestrial and freshwater plant and animal species, oceanic islands represent areas where the species can exist, surrounded by an area in which the species can survive poorly or not at all and which consequently represents a distributional barrier. Many situations that do not actually involve oceanic islands nevertheless possess the same distributional significance for many species. Thus, for alpine species a mountain top is a distributional 'island' surrounded by a 'sea' of lowlands; 129 Biol. Conserv. (7) (1975)-- © Applied Science Publishers Ltd, England, 1975 Printed in Great Britain 130 JARED M. DIAMOND for an aquatic species a lake or river is a distributional island surrounded by a sea of land; for a forest species a wooded tract is a distributional island surrounded by a sea of non-forest habitat; and for a species of the intertidal or shallow-water zones, these zones represent distributional islands compressed between seas of land and of deep water. Situation at time when protec- Original equilibrium situation tive measures go into force Final equilibrium situation 0 0 9 Fig. 1. Illustration of why the problems posed by desi$ning a system of natural reserves are similar to the problems of island biogeography. In the situation before the onset of accelerating habitat destruction by modern man, many natural habitats were present as continuous expanses covering large areas (indicated by shaded areas of sketch on left). Species characteristic of such habitats were similarly distributed over large, relatively continuous expanses. By the time that extensive habitat destruction has occurred and some of the remaining fragments are declared natural reserves, the total area occupied by the habitat and its characteristic species is much reduced (centre sketch). The area is also fragmented into isolated pieces. For many species, such distributions are unstable. Applying the lessons of modern island biogeography to these islands of natural habitat surrounded by a sea of disturbed habitat may help predict their future prospects. Throughout the world today the areas occupied by many natural habitats, and the distributional areas of many species, are undergoing two types of change (Fig. 1). First, the total area occupied by natural habitats and by species adversely affected by man is shrinking, at the expense of area occupied by man-made habitats and by species benefited by man. Second, formerly continuous natural habitats and distributional ranges of man-intolerant species are being fragmented into disjunctive pieces. If one applies the island metaphor to natural habitats and to man-intolerable species, island areas are shrinking, and large islands are being broken into archipelagos of small islands. These processes have important practical consequences for the future of natural habitats and man-intolerant species (Preston, 1962; Willis, 1974; Diamond, 1972, 1973; Terborgh, in press, a, b; Wilson & Willis, in press). Ecologists and biogeographers are gaining increasing THE ISLAND DILEMMA 131 understanding of these processes as a result of the recent scientific revolution stemming from the work of MacArthur & Wilson (1963, 1967) and MacArthur (1972). In this paper I shall explore four implications of recent biogeographic work for conservation policies: (l) The ultimate number of species that a natural reserve will save is likely to be an increasing function of the reserve's area. (2) The rate at which species go extinct in a reserve is likely to be a decreasing function of the reserve's area. (3) The relation between reserved area and probability of a species' survival is characteristically different for different species. (4) Explicit suggestions can be made for the optimal geometric design of reserves. HOW MANY SPECIES WILL SURVIVE? Let us first examine the relation between reserve area and the number of species that the reserve can hold at equilibrium. As a practical illustration of this problem, consider the fact that we surely cannot save all the rain forest of the Amazon Basin. What fraction of Amazonia must be left as rain forest to guarantee the survival of half of Amazonia's plant and animal species, and how many species will actually survive if only 1% of Amazonia can be preserved as rain forest? Numerous model 200- 100- 50 "3 j QI~IP~O'~ Long "s 20 O~,j~• • control islands 0 exploded volcanoes E 10. Z 5 Ritter 0 2 o.oot o.ol o:, ; tb ,c;o 1,ooo to,ooo to6,ooo Area (kmz) (A) Fig. 2. Example of the relation between species number and island area in an archipelago. The ordinate is the number of resident, non-marine, lowland bird species (S) on the islands of Vitiaz and Dampier Straits near New Guinea in the south-west Pacific Ocean, plotted as a function of island area (A, in km 2) on a double logarithmic scale. The points • represent relatively undisturbed islands. The straight line S = 18.9A 0"15 was fitted by least mean squares through the points for these islands. Note that species number increases regularly with island area. The two points O refer to Long and Ritter Islands, whose faunas were recently destroyed by volcanic explosions and which have not yet regained their equilibrium species number. 132 JARED M. DIAMOND systems to suggest answers to these questions are provided by distributional studies of various plant or animal groups on various archipelagos throughout the world. If one compares islands of different size but with similar habitat and in the same archipelago, the number of species S on an island is usually found to increase with island area A in a double logarithmic relation: S = SoA z (1) where So is a constant for a given species group in a given archipelago, and z usually assumes a value in the range 0.18-0.35 (Preston, 1962; MacArthur & Wilson, 1963, 1967; May, in press). A rough rule of thumb, corresponding to a z value of 0.30, is that a tenfold increase in island area means a twofold increase in the number of species. Figure 2 illustrates the species/area relation for the breeding land and freshwater bird species on the islands of the Bismarck Archipelago near New Guinea and shows that the number of bird species increases regularly with island area. If one compares islands of similar area but at different distances from the continent or large island that serves as the main source of colonisation, then one finds that the number of species on an island decreases with increasing distance. This feature is illustrated by Fig. 3, which shows that the number of bird species on Ioo]~ o. o 1 ":. ",,,, .. • .'o o.zs- ~ • ~ • • actual S _ • near-island S %o• 00° 0125- • 00625- n I I I t I ~ I u I F 0 2,000 4,000 6,000 8,000 10,000 DISTANCE (km) Fig. 3. Example of the relation between species number and island distance from the colonisation source in an island archipelago. The ordinate is the number of resident, non-marine, lowland bird species S on tropical south-west Pacific islands more than 500 km from New Guinea, divided by the number of species expected on an island of equivalent area less than 500 km from New Guinea. The expected near-island S was read off the species/area relation for such islands (Fig. 5). The abscissa is the island distance from New Guinea. Note that S decreases by a factor of 2 per 2600 km distance from New Guinea. (After Diamond, 1972.) THE ISLAND DILEMMA 133 islands of the south-west Pacific decreases by a factor of 2 for each 2600 km of distance from New Guinea. For plants or animals with weaker powers of dispersal than birds, the fall-off in species number with distance is even more rapid. Similar findings are obtained if, instead of oceanic islands, one compares habitat 'islands' within a continent or large island. For example, isolated as enclaves within the rain forest that covers most of New Guinea are two separate areas of savanna, which received most of their plant and animal species from Australia (Schodde & Calaby, 1972; Schodde & Hitchcock, 1972). The savanna which is larger and also closer to Australia supports twice as many savanna bird species as the smaller and more remote savanna (Fig. 4). Other examples are provided by mountains rising out of the 'sea' of lowlands, such as the isolated mountain ranges of Africa, South America, New Guinea and California. Thus, the number of bird species on each 'island' of alpine vegetation at high elevations in the northern Andes increases with area of alpine habitat and decreases with distance from the large alpine source area in the Andes of Ecuador (Vuilleumier, 1970).